Preliminary investigation of dose for a dedicated mammotomography system
We use a previously reported, optimized quasi-monochromatic beam technique together with unique complex acquisition trajectories made possible with a novel, dedicated cone-beam transmission computed mammotomography (CmT) system to investigate effects of low dose imaging of pendant, uncompressed breasts. Investigators have used a guideline of dose for CmT type applications as that used for dual-view mammography (4-6 mGy for average breast size). This dose is somewhat arbitrary, and it may be possible to reduce this significantly without sacrificing image quality using our quasi-monochromatic x-ray beam, 3D complex acquisition orbits, and iterative reconstruction techniques. A low-scatter acrylic resolution phantom in various media, a breast phantom with sponge and oil-filled lesions, and a cadaver breast are used to evaluate the effect of lowered dose on resolution and image artifacts. Complex saddle acquisition trajectories (necessary to overcome cone-beam distortion) are carried out for total exposures of 96, 300, and 600 mAs over 240 projections. These exposures relate approximately to 1/10 th, 1/3 rd, and 2/3 of the standard dual view mammography dose for an average sized 50% adipose/glandular breast. Iterative reconstruction uses an OSTR algorithm with 0.125 mm 3 voxels. Image artifacts increased as dose was reduced but did not appear to greatly degrade image quality except at the lowest contrast tested (1% absolute contrast). As expected, noise increased as dose was reduced. However, this did not appear to affect resolution for rods in air (high contrast), nor rods in oil (20% absolute contrast). Resolution was reduced for rods in water (1% absolute contrast) due to increased prevalence of image artifacts as well as increased noise. Breast phantom imaging of soft lesions in a highly glandular breast (6% absolute contrast) clearly yielded the 60uL and all larger volume lesions. Preliminary biological breast tissue results illustrate excellent subjective image quality at all dose levels tested. Results indicate that our quasi-monochromatic beam together with complex orbit capability and iterative reconstruction has the potential to provide sufficient image quality for practical 3D mammotomography of uncompressed breasts at significantly lower dose than dual view mammography. This is nominally a 2-fold improvement over other approaches using circular orbits and broader spectral x-ray beams. While simple image filtering (post-reconstruction smoothing) could improve noise quality, improvements in image artifact correction and scatter correction are required to more accurately determine the lower limits on dose. A contrast-detail study is also warranted with a greater variety of lesion sizes and contrasts.
